Research On Products Oil Pipeline Network Fault Diagnosis Based On Spectral Theory Of Random Matrix

Products oil play a very important role in the development of nation's economy.In order to ensure high efficiency and safety transportation,pipeline transport is the primary choice of transportation mode.Today,with the demand for oil constantly rising,products oil pipeline network is more automatic and complex and high integration than ever.Because of industry characteristics of products oil pipeline network,we must attach great importance to pipeline network safety.So,fault diagnosis is significant to the products oil pipeline network.On the basis of the study of the relevant literature,this thesis completes the following innovative work:Firstly,in the case of many devices and measurements in pipeline network,it is proposed that the products oil pipeline network model based on cyber physical system.The three-tier structure is built for pipeline network model.Physical devices and information data in pipeline network are fully considered in designing model and the model establishes the correlation between physical devices change and information.Secondly,in case of complex products oil pipeline network has the characteristics of a large number of data and a great variety of data,the method based on spectral theory of random matrix is proposed to solve this problem.In order to analyze pipeline fault from the pipeline network point of view,the centralized fuzzy decision method is proposed for pipeline network.The method could judge pipeline state accurately and find the abnormal data depend on fuzzy designed rules without reduce the dimension of data for alarm and decrease the false alarm rate.Thirdly,in order to reduce the number of data transmission on pipeline network that deliver a great deal of data frequently and improve diagnosis precision,the method of pipeline anomaly distributed detection based on element-matrix trigger mechanism is proposed.The method builds pipeline data element-matrix based on station devices data,and element-matrix is used to judge the time of data transmission.This element-matrix trigger mechanism could maximize the solution of insufficient system resources of pipeline network due to data transmission.Besides,the pressure data of adjacent station is discounted by attenuation data model considering pipeline's angle when data in the pipeline change.And the pipeline operating condition is gained by compared with the difference between real pressure data difference and model data difference.Finally,in case of products oil pipeline network with variable topological structure,a fault diagnosis method based on the graph theory model for pipeline network is proposed.Pipeline network structure is abstracted as node and the k-medoids clustering algorithm is used to group station based on the node model in order to improve the speed and the accuracy of the anomaly diagnosis.Then the diagnosis method based on eigenvalue uses clustered station data to judge pipeline state,and it could determine the reason for pipeline fault and the stations between pipeline.